Statistical analysis
Statistical analysis was made using the repeated measures ANOVA
procedure, while the weights data fitted all the required assumptions,
arthritis score data did not meet the linear and sphericity assumption.
Thus, increasing type 1 error for the model furthermore, violation for
this assumptions is not known to bias the post hoc analysis (36).
Therefore, we aggressively adjusted for multiple comparisons using the
Bonferroni adjustment. For both models, time was set as the within
subject effect while, the treatment group as the between subjects
effect. The significance level was set at 5%. Statistical analysis and
graph plotting were made with IBM Corp. Released 2015. IBM SPSS
Statistics for Windows, Version 23.0. Armonk, NY: IBM Corp. Cytokine and
anti-collagen concentration were first assessed for distribution. Since,
normal distribution was not met, non-parametric tests were chosen
throughout the statistical analysis. Comparisons of the mean
concentrations began with the Kruskal-Wallis H test and followed with
the Dunnett’s test for pair-wise comparisons. Post-hoc comparisons,
including plotted comparisons, were adjusted for multiple comparisons
following the Bonferroni procedure. Both plots and statistical analysis
were conducted using R Core Team (2013). R: A language and environment
for statistical computing. R Foundation for Statistical Computing,
Vienna, Austria. URL http://www.R-project.org/